Python/Numpy: Is there an efficient way to calculate moving averages with multiple windows?

Question

Part of a program I'm writing has code that calculates the following:

data = np.array(..........)
param = np.array(range(100)+1)
result = np.array([data[-x:].mean() for x in param])

This code is used in a giant loop so performance is crucial. It shows that the 3rd line (result = ...) takes the most time of all - I wonder if there are better ways to do this operation?

Any suggestions are appreciated!

np.array(range(100)+1) this isn't valid? Do you mean np.arange(100)+1? Also, are you really collecting the averages of the array backwards (last sample, last 2 samples, last 3 samples, etc.) — Aero Engy
– Aero Engy, Commented Dec 29, 2020 at 3:36
For the first part indeed was a typo, but yes I will be collecting backward averages.. thanks! — Destino
– Destino, Commented Dec 29, 2020 at 11:35

Frank Yellin · Accepted Answer · 2020-12-29 03:31:14Z

3

If you add 0 to the beginning of the array, and then create its cumulative sum using np.cumsum, then finding the average between indices i and j and just (my_cumsum[j] - my_cumsum[i]) / (j - i).

This should let you vastly simplify your code.

answered Dec 29, 2020 at 3:31

Frank Yellin

11.6k1 gold badge15 silver badges27 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

Ehsan · Accepted Answer · 2020-12-29 06:04:40Z

2

I think you are looking for this:

data[::-1].cumsum()[:100]/np.arange(1,101)

answered Dec 29, 2020 at 6:04

Ehsan

12.5k2 gold badges24 silver badges36 bronze badges

Collectives™ on Stack Overflow

Python/Numpy: Is there an efficient way to calculate moving averages with multiple windows?

2 Answers 2

Comments

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Your Answer

Sign up or log in

Post as a guest

Related